Have one model build and a different model review the real diff
He built forge: GPT via Codex writes the code and Claude Opus via Claude Code reviews the actual diff rather than the builder's summary, in one command.
Source
@Giamme1 on X · 2026-10-20
Written in our own words. Credit to the original author. Link checked 2026-10-20.
What they did
He built forge: GPT via Codex writes the code and Claude Opus via Claude Code reviews the actual diff rather than the builder's summary, in one command.
Result (as claimed)
Released; his point is that the model that wrote the code shouldn't grade it.
Tools they used
Cost: OpenAI Codex: free plan, paid from $8/mo; Claude Code: from $17/mo.
Set it up
Pick a contained repo or task
Start on a branch or side project where mistakes are cheap.
Install and configure
Set up OpenAI Codex, Claude Code. Read what it can run on your machine and limit permissions where possible.
Run it on a real task
Give it a well-scoped issue and let it work. Review the diff like any pull request.
Add guardrails
Keep tests and CI in the loop, and never let an agent push to main or touch production secrets unattended.
Copy the full setup prompt
Paste into your agent and it will walk you through the setup, asking before anything that sends, posts or spends.
You are helping me set up this AI use case end to end in Claude Code. Goal: Have one model build and a different model review the real diff What the original author did: He built forge: GPT via Codex writes the code and Claude Opus via Claude Code reviews the actual diff rather than the builder's summary, in one command. Reported result (as claimed by the author): Released; his point is that the model that wrote the code shouldn't grade it. Source: https://x.com/Giamme1/status/2112567652757491726 Tools they used: OpenAI Codex, Claude Code Category: Dev & ops Please: 1. Ask me 3 short questions about my situation (accounts I have, budget, what I want the output to look like). 2. Propose a step-by-step plan using these tools or cheaper alternatives, with the time and cost of each step. Use official pricing pages; say "check pricing" if unsure. 3. Then do the setup with me one step at a time, waiting for my OK before anything that sends messages, posts publicly, spends money or deletes data. 4. Finish with a checklist to run it again next time.
You are helping me set up this AI use case end to end in Codex. Goal: Have one model build and a different model review the real diff What the original author did: He built forge: GPT via Codex writes the code and Claude Opus via Claude Code reviews the actual diff rather than the builder's summary, in one command. Reported result (as claimed by the author): Released; his point is that the model that wrote the code shouldn't grade it. Source: https://x.com/Giamme1/status/2112567652757491726 Tools they used: OpenAI Codex, Claude Code Category: Dev & ops Please: 1. Ask me 3 short questions about my situation (accounts I have, budget, what I want the output to look like). 2. Propose a step-by-step plan using these tools or cheaper alternatives, with the time and cost of each step. Use official pricing pages; say "check pricing" if unsure. 3. Then do the setup with me one step at a time, waiting for my OK before anything that sends messages, posts publicly, spends money or deletes data. 4. Finish with a checklist to run it again next time.
You are helping me set up this AI use case end to end in Grok Bot. Goal: Have one model build and a different model review the real diff What the original author did: He built forge: GPT via Codex writes the code and Claude Opus via Claude Code reviews the actual diff rather than the builder's summary, in one command. Reported result (as claimed by the author): Released; his point is that the model that wrote the code shouldn't grade it. Source: https://x.com/Giamme1/status/2112567652757491726 Tools they used: OpenAI Codex, Claude Code Category: Dev & ops Please: 1. Ask me 3 short questions about my situation (accounts I have, budget, what I want the output to look like). 2. Propose a step-by-step plan using these tools or cheaper alternatives, with the time and cost of each step. Use official pricing pages; say "check pricing" if unsure. 3. Then do the setup with me one step at a time, waiting for my OK before anything that sends messages, posts publicly, spends money or deletes data. 4. Finish with a checklist to run it again next time.
The Daily 5
Get five new use cases tomorrow morning
Free. Real sources. Unsubscribe any time.
Guides for this use case
How to vibe code an app with Claude Code · How to use Claude Code and Codex together · How to run multiple coding agents together
Related use cases
Have a second AI review your agent-built app as if it must handle 10x the data
The footage search Claude Code built him looked done, until a second AI found every vector held in RAM, 8 to 10 GB at 1,000 hours of footage by Claude's estimate. His review prompt: list each problem as a numbered fix with the file, why and the exact change.
Share and review agent skills with teammates who do not use Git
Skill Dockyard helps teams share and review Claude Code and Codex skills, so non-developers can use the team's skills without managing them in GitHub.
Let an agent review and fix every open PR overnight, and watch your usage
A developer let Amp work overnight on his open-source repo with one prompt: review every PR, wait for CI and Codex bot review comments, resolve them, re-request review, run an end-to-end test and prove everything worked. Claude Code hit its usage limit for 5 hours mid-run.
Review pull requests by what the code does, not only the diff
Built Perspica, a semantic diff that explains what changes do and the concepts behind them for AI-era code review.